sjh-skills
AGENTS.md43 starsChanged 7 months agoMITView on GitHub
A collection of Claude Code skills for research workflow automation. Each skill is self-contained under skills/<name>/ with its own SKILL.md, scripts/, and references/. Every skill follows this structure: All install commands must use the shorthand format: Never use full https://github.com/... URLs or point to standalone repos that no longer exist.
Where files are headingNew AGENTS.md files have more rules about keeping secrets out of the code, and fewer about code style.
Of the AGENTS.md files we read, the 32,671 first seen in the last six months, against the older ones: the kind of rule that gained most share, and the kind that lost most. About files like this in general, not this one, and an estimate, since a kind is a pattern in the words.
In this repository17 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- bibtex-fetchSkill
- context-auditSkill
- daily-summarySkill
- experiment-registrySkill
- handoffSkill
- init-projectSkill
- notion-lifeosSkill
- paper-analyzerSkill
- paper-self-reviewSkill
- project-reviewSkill
- remote-cluster-agentSkill
- research-leadSkill
- session-teamSkill
- sync-docsSkill
- team-digestSkill
- web-fetcherSkill
The file
We split this AGENTS.md into the rules it gives an agent
11 rules. About 0 are common, 3 are distinctive, and 8 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈3
Distinctive · estimate
This author's own choices. Worth a look.
8
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 38
Every skill follows this structure:
- Distinctive: line 58
All install commands must use the shorthand format:
- Very specific: line 64
Never use full https://github.com/...
- Very specific: line 70
Read the existing SKILL.md before modifying a skill
- Very specific: line 71
Update CHANGELOG.md for every change — group entries under the skill name, use Keep a Changelog format
- Distinctive: line 73
Keep install commands consistent across all READMEs (root, English, Chinese) when adding or renaming a skill
- Very specific: line 74
Update the skill table in both README.md and README.zh-CN.md when adding a new skill
- Very specific: line 75
Test skill triggering by reviewing the description field — it's the primary mechanism Claude uses to decide whether to invoke a skill
- Very specific: line 87
Write SKILL.md body in Chinese (description trigger words are fine)
- Very specific: line 88
Use README.zh.md naming — always use README.zh-CN.md
- Very specific: line 89
Use full GitHub URLs in install commands — use jiahao-shao1/sjh-skills --skill <name>
≈ “Common” is an estimate: at least 1 in 5 of the 41,581 AGENTS.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
SJH Skills
A collection of Claude Code skills for research workflow automation. Each skill is self-contained under skills/<name>/ with its own SKILL.md, scripts/, and references/.
Repository Structure
sjh_skills/
├── skills/
│ ├── daily-summary/ # Git + Claude sessions + Notion aggregation
│ ├── notion-lifeos/ # PARA method + Make Time via Notion API
│ ├── web-fetcher/ # 5-layer fallback web content extraction
│ ├── init-project/ # Claude Code project initialization
│ ├── project-review/ # 5-dimension strategy review snapshot
│ ├── remote-cluster-agent/ # Remote GPU cluster ops via Go daemon + rca CLI
│ ├── paper-analyzer/ # Deep critical paper analysis with causal chain methodology
│ ├── paper-self-review/ # Paper paragraph self-review (logic / repetition / detail leakage)
│ ├── experiment-registry/ # ML experiment registry with YAML + CLI
│ ├── handoff/ # Session handoff summary for context continuity
│ ├── session-team/ # Leader / worker teams of sessions on one machine
│ ├── team-digest/ # Team communication through agents' commits and topic issues
│ ├── sync-docs/ # Documentation sync — updates docs and commits
│ ├── context-audit/ # Progressive disclosure compliance audit
│ └── bibtex-fetch/ # arXiv / Semantic Scholar BibTeX fetcher
├── packages/ # Pip-installable packages (e.g. exp-registry)
├── hooks/ # Optional hook helpers; no plugin hooks are auto-registered by default
├── .claude-plugin/ # Plugin manifest and marketplace config
├── .codex/ # Codex CLI install instructions
├── CHANGELOG.md # Version history (must be updated on every change)
├── README.md # English
└── README.zh-CN.md # ChineseConventions
Skill Anatomy
Every skill follows this structure:
skill-name/
├── SKILL.md # Required — frontmatter (name, description) + instructions
├── README.md # English documentation
├── README.zh-CN.md # Chinese documentation (if applicable)
├── scripts/ # Executable scripts
├── references/ # Docs loaded into context as needed
└── assets/ # Templates, icons, fontsLanguage
- SKILL.md — Written in English. Chinese trigger words preserved in the
descriptionfield. - README.md — English.
- README.zh-CN.md — Chinese (use
zh-CNsuffix, notzh).
Install Commands
All install commands must use the shorthand format:
npx skills add jiahao-shao1/sjh-skills --skill <name>Never use full https://github.com/... URLs or point to standalone repos that no longer exist.
Behavior Boundaries
Always Do
- Read the existing SKILL.md before modifying a skill
- Update
CHANGELOG.mdfor every change — group entries under the skill name, use Keep a Changelog format - Bump
.claude-plugin/plugin.jsonversion on every skill change — minor for new features, patch for fixes/improvements - Keep install commands consistent across all READMEs (root, English, Chinese) when adding or renaming a skill
- Update the skill table in both
README.mdandREADME.zh-CN.mdwhen adding a new skill - Test skill triggering by reviewing the
descriptionfield — it's the primary mechanism Claude uses to decide whether to invoke a skill
Ask First
- Renaming a skill directory (breaks existing installs)
- Removing a skill from the monorepo
- Changing the
descriptionfield format or triggering strategy - Adding external dependencies to a skill's
scripts/
Never Do
- Hardcode API keys, tokens, or user-specific paths in SKILL.md or scripts
- Write SKILL.md body in Chinese (description trigger words are fine)
- Use
README.zh.mdnaming — always useREADME.zh-CN.md - Use full GitHub URLs in install commands — use
jiahao-shao1/sjh-skills --skill <name> - Point install commands to standalone repos (e.g.,
jiahao-shao1/notion-lifeos-skill) — everything is in the monorepo now
Changelog
CHANGELOG.md follows Keep a Changelog. Every change must be recorded before committing:
- Added — new skills, features, scripts
- Changed — rewrites, refactors, behavior changes
- Fixed — bug fixes
- Improved — optimization, description tuning
- Removed — deleted features or skills
Bump the version number when releasing a coherent set of changes. Use semver: major (breaking), minor (new skill or feature), patch (fixes and improvements).
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

